Purpose: We improve the accuracy of Twitter sentiment analysis with a hybrid model combining Word to Vector (Word2Vec) and the Robustly Optimized BERT Pretraining Approach (RoBERTa). The idea is that Word2Vec is strong for slang/novel vocabulary (distributional semantics), while RoBERTa excels in contextual meaning; combining the two mitigates each other's weaknesses. Methods/Study design/approach: The Sentiment140 dataset contains 1.6 million balanced tweets. The split is stratified; Word2Vec is trained solely on the training data. RoBERTa is pretrained (frozen in the first stage, then fine-tuned with some layers in the second stage). The Word2Vec and RoBERTa vectors are concatenated and processed using Bidirectional Long Short-Term Memory (BiLSTM) with sigmoid activation. Training utilizes TensorFlow and the Adam optimizer, incorporating dropout and early stopping. The decision threshold is optimized during the validation process. The process supports caching and training resumes. Result/Findings: The hybrid model achieved an accuracy of 88.09%, an F1-score of 88.09 %, and an Area Under the Curve (AUC) ≈ 95.19% on the Receiver Operating Characteristic (ROC). No overfitting was observed, and the hybrid model outperformed both single baselines. The confusion matrix and ROC curve corroborate the findings. Novelty/Originality/Value: The novelty lies in the fusion of distributional and contextual representations with resource-efficient fine-tuning. Limitations: Computational requirements and hyperparameter tuning are not yet extensive. Further directions: systematic hyperparameter search and cross-validation across other large sentiment datasets to assess generalization.

Outpatient diagnoses in primary care: Adding transparency in the field of outpatient diagnoses

This perspective paper explores the bidirectional influence between language emergence and the relational structure of subjective experiences, termed qualia structure, and lays out a constructive approach to the intricate dependency between the two. We hypothesize that the emergence of languages with distributional semantics (e.g., syntactic-semantic structures) is linked to the coordination of internal representations shaped by experience, potentially facilitating more structured language through reciprocal influence. This hypothesized mutual dependency connects to recent advancements in AI and symbol emergence robotics, and is explored within this paper through theoretical frameworks such as the collective predictive coding. Computational studies show that neural network-based language models form systematically structured internal representations, and multimodal language models can share representations between language and perceptual information. This perspective suggests that language emergence serves not only as a mechanism for creating a communication tool but also as a mechanism for allowing people to realize shared understanding of qualitative experiences. The paper discusses the implications of this bidirectional influence in the context of consciousness studies, linguistics, and cognitive science, and outlines future constructive research directions to further explore this dynamic relationship between language emergence and qualia structure.

The translation of the Babylonian Talmud into Italian constitutes a true challenge. To address the numerous issues posed by translating such a vast and complex text, the Traduco system was developed to support translators through collaborative features, a translation memory, distributional semantics, glossary management, and annotation tools. Additionally, recent experiments have explored the integration of large language models to assist in the translation and the construction of a Talmud Knowledge Base designed to provide a formal representation of the terminological and conceptual data identified in the Talmud.

ABSTRACT Contextual word embeddings have proven valuable for analysing word meanings, yet their application to Chinese polysemy remains underexplored. This study examines how polysemous and monosemous Chinese words behave in embedding space using the Chinese-ROBERTA-wwm-ext model. We analyse 59 words from the Chinese Wikipedia corpus through theoretically-grounded metrics based on information theory and distributional semantics, including embedding magnitude (information content) and standard deviation (contextual variability). Our findings reveal no statistically significant differences between polysemous and monosemous words in their distributional patterns, providing crucial cross-linguistic validation of previous findings in English. This cross‑linguistic consistency suggests a potentially language‑independent principle in how contextual embeddings encode lexical meaning, where meaning variation exists on a continuous spectrum rather than in discrete categories. This aligns with similar findings in English, indicating that the relationship between dictionary-defined meanings and distributional semantics might be fundamentally more complex than previously theorized.

Abstract This paper offers some observations on the learning of argument structure constructions as well as recommendations for teaching argument structure constructions to intermediate to advanced learners, notably argument structure constructions that could be considered to alternate such as the intransitive non-causative construction (INCCx) and the transitive causative construction (TCCx). Our aim is to show that for such schematic argument constructions, generalisations emerge from the interaction between the verb, its arguments and the construction(s) they occur in. Through a distributional semantics analysis of the themes (the argument in subject position in the INCCx and object position in the TCCx), we start by identifying constructional meaning for each construction. Then, through computational simulations of learning we identify which elements in the constructions are the most reliable cues for learners to choose one or the other construction. This work is inspired by constructional approaches to language (e.g., Goldberg, 1995 ) and learning theory (for linguistics, e.g., Skinner, 1957 ). Through our constructional approach and learning simulations, we identify characteristics for each construction that can be used to teach learners the use of each construction generally. We also observe that the TCCx appears easier to learn and potentially more productive than the INCCx and provide examples of pedagogical materials.

Abstract This paper proves a homomorphism between extensional formal semantics and distributional vector space semantics, demonstrating structural compatibility. Formal semantics models meaning as reference, using logical structures to map linguistic expressions to truth conditions, while distributional semantics represents meaning through word vectors derived from contextual usage. By constructing injective mappings that preserve semantic relationships, we show that every semantic function in an extensional model corresponds to a compatible vector space operation. This result respects compositionality and extends to function compositions, constant interpretations, and n -ary relations. Rather than pursuing unification, we highlight a mathematical foundation for hybrid cognitive models that integrate symbolic and sub-symbolic reasoning and semantics. These findings support multimodal language processing, aligning ‘meaning as reference’ (Frege, Tarski) with ‘meaning as use’ (Wittgenstein, Firth).

In the field of forensic science, mixed DNA evidence obtained from crime scenes often contains genetic information from multiple individuals, and its accurate interpretation is crucial for case investigation and judicial decision-making. With the advancement of forensic genetic technologies, although detection capabilities have significantly improved, there are still substantial bottlenecks in the interpretation of multi-contributor DNA profiles. Traditional methods are often unable to simultaneously and precisely infer both the genotypes of suspects and their respective contribution proportions, which makes them insufficient to meet the stringent requirements of complex mixture analysis. To address these challenges, we propose a continuous gamma distribution model algorithm based on probabilistic residual optimization in this study. By constructing a two-step probabilistic evaluation framework, the algorithm first generates candidate genotype combinations through allelic permutations and estimates preliminary contributor proportions. It then introduces the gamma distribution hypothesis to build a probability density function, dynamically optimizes the shape parameter (α) and the scale parameter (β) to calculate residual probability weights, and employs an iterative maximum likelihood estimation process to simultaneously optimize genotype combinations and contributor proportion parameters. The final results are derived by integrating population allele frequency databases to output the maximum likelihood solution. This algorithm provides a reliable and quantifiable analytical tool for forensic identification, significantly improving the accuracy of complex mixture interpretation and enhancing the practical utility of mixed DNA in criminal investigations. It holds substantial significance in advancing forensic science technologies and safeguarding judicial fairness.

MotivationHypothesis generation is a fundamental problem in biomedical text mining that aims to generate ideas that are new, interesting, and plausible by discovering unexplored links between biomedical concepts. Despite significant advances made by existing approaches, they do not fully leverage the evolutionary properties of biomedical concepts. This is limiting because scientific knowledge continually evolves over time, with new facts being added and old ones becoming obsolete. Thus, it is crucial to capture the evolutionary properties of biomedical concepts from multiple perspectives (e.g. spatial, temporal, and semantic) to generate hypotheses that reflect the up-to-date information landscape of the biomedical domain.ResultsWe introduce a novel framework, ConceptDrift, that models the hypothesis generation task as a sequence of temporal graphlets and simultaneously encodes spatial, temporal, and semantic change. Unlike existing approaches that treat these dimensions independently, ConceptDrift is the first to provide a holistic understanding of concept evolution by integrating them into a unified framework. Grounded in the theories of the Distributional Hypothesis and Conceptual Change, our method adapts these principles to the unique challenges of large-scale biomedical literature. We conduct extensive experiments across multiple datasets and demonstrate that ConceptDrift consistently outperforms state-of-the-art baselines in generating accurate and meaningful hypotheses. Our framework shows immediate practical benefits for web-based literature mining tools in life sciences and biomedicine, offering more robust and predictive feature representations.Availability and implementationhttps://github.com/amir-hassan25/ConceptDrift (DOI: 10.6084/m9.figshare.29975476).

This study presents a consistent and transferable methodology for identifying and analyzing continuous urban electric traction networks—specifically trams, trolleybuses, and metro systems—in Central European cities. Using data from OpenStreetMap, enriched with official transport sources and field verification, we defined 113 integrated networks across Czechia, Hungary, Germany, Poland, Austria, Slovakia, and Switzerland. Key absolute and relative indicators—such as total network length and composite density—were calculated for each system, and linked to standardized statistical units (LAU, NUTS). At least minimal electric public transport infrastructure was identified in 366 municipalities. The study examines spatial patterns in network distribution and tests several hypotheses, including whether capital cities with metro systems also host the most extensive tram and trolleybus networks, and how urbanization levels affect network density. The findings confirm that larger and more urbanized cities tend to support more complex and denser electric transport systems, but also reveal exceptions influenced by historical and spatial factors. The analysis demonstrates that using only urbanized areas yields more meaningful comparisons than relying on entire administrative boundaries. A seven-category typology was developed to enable comparative assessment of network significance and urban transport potential across the region. The results offer a robust database for further spatial and transport analyses and highlight the value of network density as an indicator of public transport quality. This approach can be applied in other regions worldwide, supporting sustainable mobility research and planning.

The dwarf red brocket, Mazama rufina (Pucheran, 1851) is a small deer with a fragmented distribution in the montane forests of the Andes of Peru, Ecuador, Colombia, and Venezuela. Little is known about the phylogenetic relationships and the haplotype diversity of its populations, which show distribution gaps. Here we elucidate the phylogenetic relationships of M. rufina and other neotropical deer using mitochondrial data, and analyze genetic geographic variation of this taxon by using haplotype networks of the Cyt-b gene from northern South America. Our analyses recovered M. rufina as independent clade that is not part of Mazama, and sister to a clade composed of Mazama species (except Mazama chunyi) and Odocoileus. The morphometric data of cranial traits confirms that the dwarf red brocket is among the smallest species of deer in South America, only overlapping with small cis-Andean gray brockets (genus Passalites). Based on these results, we provide a new generic classification for this taxon by placing the dwarf red brocket in a new genus found only in the Andes of northern South America. The Cyt-b haplotype network of the dwarf red brocket showed a strong geographic structure caused by the interplay of Cordilleras and lowland river valleys. The genetic distances between the geographic groups were between 1.4 % (Central Cordillera of Colombia vs. Andes of Ecuador) to 2.52 % (Mrida Cordillera vs. Ecuador). The species range using Extent of Occurrence (EOO) and Area of Occupancy (AOO) was 443,764 and 796 km2, respectively, suggesting that the species could be listed as Near Threatened. However, additional information on population changes and susceptibility to habitat transformation is crucial to evaluate whether the dwarf red brocket can be deemed Vulnerable along its distribution. Compared with previous distribution hypotheses, the revised map suggests less extensive distribution gaps in Colombia and highlights priority areas for future sampling in Colombia, Ecuador, and Venezuela.

Vector Space Models (VSM) and neural word embeddings are core components in recent Machine Learning (ML) and Natural Language Processing (NLP) pipelines. By encoding words, sentences and documents as high-dimensional vectors via distributional semantics, they enable Information Retrieval (IR) systems to capture semantic relatedness between queries and answers. This paper compares different semantic representation strategies for query-statement matching, evaluating paraphrase identification within an IR framework using partial and syntactically varied queries of different lengths. Motivated by the Word Mover’s Distance (WMD) model, similarity is evaluated using the distance between individual words of queries and statements, as opposed to the common similarity measure of centroids of neural word embeddings. Results from ranked query and response statements demonstrate significant gains in accuracy using the combined approach of similarity ranking through WMD with the word embedding techniques. Our top-performing WMD + GloVe system consistently outperformed Doc2Vec and an LSA baseline across three return-rate thresholds, achieving 100% correct matches within the top-3 ranked results and 89.83% top-1 accuracy. Beyond the substantial gains from WMD-based similarity ranking, our results indicate that large, pre-trained word embeddings, trained on vast amounts of data, result in portable, domain-agnostic language processing solutions suitable for diverse business use cases.

A map of words: Retrieving the spatial layout of medium-scale geographical maps through distributional semantics.

‎In practice‎, ‎we may encounter hypotheses that the parameters under test have typical restrictions‎. ‎These restrictions can be placed in the null or alternative hypotheses‎. ‎In such a case‎, ‎the hypothesis is not included in the classical hypothesis testing framework‎. ‎Therefore‎, ‎statisticians are looking for the more powerful tests‎, ‎rather than the most powerful tests‎. ‎A common method for such tests is to use intersection-union and union-intersection tests‎. ‎In this paper‎, ‎we derived the testing procedure of a simple intersection-union and compared it with the likelihood ratio test‎. ‎We also compare the powers of two exponential sign tests‎, ‎the rectangle test and smoother test‎, ‎and the simple intersection-union test with the likelihood ratio test‎.

ABSTRACT NFT market is nascent and thus prone to manipulative behavior. This paper examines the impact of wash trading on the relationships between NFT returns, volume, and volatility via Mixture of Distributions Hypothesis (MDH) and Sequential Information Arrival Hypothesis (SIAH), and the role of collection characteristics in these dynamics via Hedonic Pricing Theory (HPT). By comparing the full dataset and those devoid of cyclical wash trades, we find that MDH and SIAH hold across samples. Notably, the return-volatility relationship shifts from significantly negative to significantly positive post-cleaning, confirming that manipulative trades distort true market risk-return dynamics. In contrast, support for HPT weakens after applying stricter wash trade detection, suggesting collection features had overstated influence due to manipulation. These findings highlight the need for robust wash trading detection to ensure data reliability. Policymakers should consider ensuring market data reliability by enhancing transparency regulations around suspected wash trade transactions.

Active West Nile virus transmission in Brazil: an epidemiological study

Abstract Data-driven neural word embeddings (NWEs), grounded in distributional semantics, can capture various ranges of linguistic regularities, which can be further enriched by incorporating structured knowledge resources. This work proposes a novel post-processing approach for injecting semantic relationships into the vector space of both static and contextualized NWEs. Current solutions to retrofitting (RF) word embeddings often oversimplify the integration of semantic knowledge, neglecting the nuanced differences between relationships, which may result in suboptimal performance. Instead of applying multi-thresholding to distance boundaries in metric learning, we compute taxonomic similarity to dynamically adjust these boundaries during the semantic specialization of word embeddings. Benchmark evaluations on both static and contextualized word embeddings demonstrate that our dynamic-fitting (DF) approach produces SOTA correlation results of 0.78 and 0.76 on SimLex-999 and SimVerb-3500, respectively, highlighting the effectiveness of incorporating multiple semantic relationships in refining vector semantics. Our approach also outperforms existing RF methods in both supervised and unsupervised semantic relationships recognition tasks. It achieves top accuracy scores for hypernymy detection on the BLESS, WBLESS, and BIBLESS datasets (0.97, 0.89, and 0.83, respectively) and an F1 score of over 0.60 on four types of semantic relationship classification in the shared Subtask-2 of CogALex-V, surpassing all participant systems. In the analogy reasoning task of the Bigger Analogy Test Set, our approach outperforms existing RF methods on inferring relational similarity. These consistent improvements across various lexical semantics tasks suggest that our DF approach can effectively integrate distributional semantics with symbolic knowledge resources, thereby enhancing the representation capacity of word embeddings in downstream applications.

Aim: The COVID-19 pandemic is an emergent viral respiratory disease characterized by high fever and shortness of breath, and it was declared a pandemic by the World Health Organization in March 2020. Early assessment of patients’ biochemical tests is important for accelerating diagnosis, allowing effective treatment, and controlling the further spread of the disease. The present study aimed to investigate the association between the disease, trace elements -including copper (Cu), zinc (Zn), selenium (Se), manganese (Mn), and cobalt (Co) vitamin D, Alanine aminotransferase (ALT) and Aspartate aminotransferase (AST) biochemical levels, and the correlation between the parameters tested in patients with COVID-19. Methods: In our study, 40 patients (case group) who were hospitalized with a diagnosis of COVID-19 based on chest X-ray images and RT-PCR results evaluated by an infectious diseases specialist were included, along with 40 healthy individuals (control group) over the age of 18 who had no prior symptoms of COVID-19, no visits to a medical doctor for COVID-19, and no history of hospitalization due to the disease. Beckman Coulter AU5800 (Beckman Coulter, Brea, CA, USA) autoanalyzer was used for spectrophotometric analyses of clinical biochemistry tests, and vitamin D levels were examined using the HPLC method with the Shimadzu SIL-20A HT autosampler. Levels of trace elements-including Cu, Zn, Se, Mn, and Co-were measured by inductively coupled plasma mass spectrometry (ICP-MS) on an ICP-MS Bruker Aurora M90 analytical complex. The normal distribution hypothesis for the variables in question was tested using the Kolmogorov–Smirnov test. Student’s t-test was used for intergroup comparisons of variables meeting the normal distribution hypothesis, whereas Mann–Whitney U test was used for variables that did not meet the hypothesis. Results: Vitamin D levels were much lower in the case group (12.05 ng/mL ± 6.27) compared to the control group (23.54 ng/mL ± 10.54), and the difference was statistically significant (p < 0.001). Serum Cu, Zn, Se, Mn, and Co levels in the control group were higher compared to the COVID-19 group, yet only the differences in Zn, Se, and Mn levels were statistically significant (p

Introduction. The Bethesda System for Reporting Thyroid Cytopathology allows to standardize interpretation of fine needle aspiration biopsy of thyroid nodules with assessment of suspected malignancy risk. Analysis of prognostic significance of cytological data, using the Bethesda System as an example, is necessary for determination of tumor process aggressiveness and possibility of development of recurrences.Aim. To study prognostic significance of the Bethesda System for Reporting Thyroid Cytopathology in recurrent thyroid carcinoma.Materials and methods. Results of examinations and surgical interventions performed at the Altai Regional Oncological Dispensary (Barnaul) during a 3-year period in 503 patients (423 women (84.1 %) and 80 (15.9 %) men) with thyroid carcinoma are analyzed. In 440 patients, thyroid carcinoma in surgical material was found for the first time (R0 – primary patients); in 63 patients, disease recurrence was diagnosed (R1 – presence of recurrence). At the time of surgery, median patient age was 51 years (interquartile range 40–62 years). The inclusion criterion for the study was morphologically confirmed diagnosis of thyroid carcinoma; the exclusion criterion was the absence of cytological examination per the Bethesda System in patients with recurrent carcinoma. Data from medical records, disease histories and cancer registry were evaluated taking into account clinical and anamnestic characteristics of a special classifier for determination of risk factors of disease recurrence. Data processing was performed using Microsoft Excel software. At the preoperative stage, in primary patients, results of ultrasound and cytological examination of the fine needle aspiration biopsy of thyroid nodules were analyzed. Results of cytological examination per the Bethesda System were considered as an independent risk factor of recurrence. Orange Data Mining (version 3.3.37.0) and RStudio (version 4.3.1) software were used for data analysis. Distributions of quantitative variables were evaluated using the Kolmogorov–Smirnov and Shapiro–Wilk tests. Normal distribution hypothesis was rejected at p <0.20 for Kolmogorov–Smirnov test and at p <0.05 for Shapiro–Wilk test.Results. According to the data, median time to recurrence was 2 years (Q1–Q3 – 1–6 years), minimal time to its development was 1 year, maximal – 20 years. 1-year recurrence-free survival was 100 %, 3-year was 91 %, 5-year was 86 %, 10-year was 59 %, 15-year was 25 %. Mean time to recurrence in 50 % of patients was 9.8 ± 0.9 years. Risk of recurrence of thyroid carcinoma of category IV per the Bethesda System was 3.623 times (72.4 %) lower compared to category III (hazard ratio 0.276; 95 % confidence interval 0.110–0.691; p = 0.006). Recurrence-free survival in patients with thyroid carcinoma of category VI per the Bethesda System was significantly higher (р = 0003 and p < 0.001) than in patients with category III and IV cytopathology.Conclusion. Cytological conclusion of category III per the Bethesda System is 2.755 times more common in recurrence group than of category VI. Category VI cytopathology, 88.7 % of which are papillary thyroid carcinoma with conclusive cell signs, is characterized by relatively low risk of recurrence. Cytologic conclusion of categories III and IV per the Bethesda System is an unfavorable prognostic factor in preoperative diagnostics due to undetermined interpretation of the pathology. These categories are commonly observed in patients with more aggressive disease progression.

One of the four most central corpus-linguistic methods is keywords/keyness analysis, which is generally the identification and interpretation of word types of a target corpus ( T) that, when compared to their occurrence in a reference corpus ( R), are key/characteristic for T. In this article, I will (a) apply methods proposed by Gries (2021) to the study of three outer-circle varieties of English to identify cultural keywords in a bottom-up fashion and (b) use the results from that first application to advance two suggestions how to extend keyness analyses to better understand the keywords from the first step: key collocates, which involves applying keyness to contexts of keywords; and deep key collocates, which involves distributional semantics methods like word2vec, GloVe, BERT, etc. to keywords. I will use Mukherjee and Bernaisch's (2015) keyness analysis as a launchpad to identify keywords from comparisons of Indian, Pakistani, and Sri Lankan Englishes (IndE, PakE, and SriE, respectively) and zoom in on the variety-specific differences of the keyword of terror. The results not only indicate what terms are key for which of the three varieties; they also allow for a new level of granularity in how keywords use differs across varieties and possibly cultures. For example, in the PakE data, newspaper coverage of terror is mostly discussed with regard its financial aspects and implications and matters of communication, whereas in IndE and SriE, terror is much more approached from a military and a religious perspective, respectively. 1